Const-me
Comparing to whisper-ctranslate2
When I tried whisper-qtranslate 2, I was surprised by the results of audio recognition-It`s clearly better, fewer mistakes and almost no repetitions of echoes of the same words. Please take a closer look at this fork. And use their achievements.
https://github.com/jordimas/whisper-ctranslate2
nice from you to spread the info about other projects, always interesting. Sadly just like the original whisper project, it just don't work (for me)... the python stuff is just dependency hell...
Ctranslate2 is really great, I used it on Japanese. I think the default beam size of 5 yields better quality than Const-me. I had to change -device CPU as CUDA didn't work here so it's very slow.
